Guide
What changes from prototype to production on a machined part, and why do parts that passed as prototypes fail in production?
Prototype to production is the move from making a few parts to prove a design to making repeated lots with a frozen drawing, a fixed process route, dedicated fixtures, an agreed inspection plan and a unit cost the program can live with. Parts that passed as prototypes fail in production when something that made the prototype good was never written down: a machinist’s careful fitting, a dimension checked against the CAD model instead of a tolerance, a material or finish the prototype did not have, or a sample too small to show how much the real process varies.
By Saola Manufacturing engineering team · Published · Updated
The story is familiar to design engineers and buyers alike. A handful of prototypes were machined, they fit, the test rig was happy and the design was released. Then the first production lot arrives and a bore position fails incoming inspection, threads bind after anodizing, or the parts measure in tolerance but will not assemble. Or nothing fails, but the production quote is far above what the prototype price suggested. In each case the design is usually not the problem. The prototype proved that the geometry can work; it did not prove that a repeatable process, run by different people on different machines with different fixtures, will keep making it. This guide lists the causes, shows how to find the one you have, gives a change checklist for process, fixturing, first article and drawing cleanup, and lists what to put on the drawing, RFQ and purchase order so the next lot does not repeat it.
On this page
How do you recognize a prototype-to-production problem on a machined part?
The signature is a part that was accepted as a prototype and then fails at first article, at incoming inspection or on the assembly line once production lots arrive, while the drawing has not changed. A second signature is a production quote that does not come down with quantity the way the team expected.
Look for these patterns before calling it a supplier quality problem:
- Features that relate two setups fail first. Position of a hole pattern relative to a turned diameter, perpendicularity of a face milled in a second operation, or coaxiality between bores machined from opposite sides. These depend on fixturing and datums, which are exactly what changes between a one-off prototype and a production routing.
- Parts are in tolerance but do not assemble. The prototypes were fitted to their mating parts, or the mating parts were themselves prototypes, so the tolerance stack was never tested at its limits.
- Problems appear only after finishing. The prototypes were left bare or given a quick finish, while production parts are anodized, plated or coated, which adds thickness to threads, bores and fits.
- The lot is mostly good but some parts are not. Three prototypes cannot show the spread of a process; a lot of hundreds will.
- Supplier and customer measurements disagree. The prototype was checked with calipers or against the model on a CMM, and production is checked with gauges or a different datum setup.
- The cost does not scale. The prototype shop absorbed hand deburring, careful setups and loose interpretation of unclear notes; a production quote has to price every requirement as written, plus fixtures, gauges and documents.
Why do parts that passed as prototypes fail in production?
In the order worth checking them, the usual causes are an incomplete drawing that the prototype shop filled in by judgment, a sample too small to show process variation, a different process route and fixturing in production, a material or finish change, and a different way of measuring.
- The drawing did not define the part; the prototype maker did. Prototypes are often cut straight from a 3D model, with the machinist deciding which faces matter. Untoleranced dimensions, missing datums and notes such as “match mating part” are resolved by skill and conversation. ASME Y14.5 sets the opposite expectation: every dimension has a tolerance, and the drawing defines the part without depending on how it is made. A production supplier works to what is written, so the gaps become rejects or disputes.
- A few good parts say nothing about the spread. Prototypes show that the design can be made once. The NIST/SEMATECH e-Handbook defines capability against an in-control process and puts a large enough sample for a capability index at about 50 independent values; three or five prototypes are far short of that. A tolerance that a prototype machinist could hold by measuring and recutting may be outside what the production process holds every time.
- Production uses a different process route. A prototype may be milled from solid in one long five-axis setup with probing between cuts. Production may split the same part into two or three operations on dedicated fixtures to reach a sensible cycle time. Each new setup adds a relocation error, so features that were cut together in the prototype are now cut in different clampings.
- The material, stock form or blank changed. Plate instead of bar, a different temper, a different mill or a cast or forged blank instead of solid stock all change residual stress, distortion after unclamping and machinability. Thin walls that stayed flat in a prototype can move once a different blank is roughed in a single pass.
- A finish was added or changed. Anodize, plating and coatings build up on every surface they cover, including threads and bores, unless the drawing says what is masked and whether dimensions apply before or after finishing. The threads after finishing guide shows how that buildup closes a thread.
- The measurement method changed. Calipers, a CMM program, hard gauges and a functional fixture can give different answers on the same part, especially for form and position tolerances. If the prototype and production checks are not the same method, the disagreement may be in the measurement, not the part.
Cost surprises follow the same logic. A prototype price covers programming and setup spread over a few pieces, but it also hides work that the prototype shop did informally. A production quote adds fixtures, gauges, a first article report, finishing and packaging, and prices every tolerance as written. The CNC machining cost drivers guide and the NRE guide show how those one-time and per-part lines are built.
How do you diagnose why production parts fail when the prototypes passed?
Compare the prototype and production records feature by feature: which characteristics fail, how each was made and how each was measured. The pattern of failing features usually points to one cause before anyone needs to argue about blame.
- List the failing characteristics by balloon number. Use a ballooned drawing so each rejected dimension has an identity; the ballooned drawing guide explains the method. Note whether the failures are on one feature, on features that relate two setups, or scattered.
- Find the prototype evidence for the same characteristics. Often there is none: the prototype was accepted on fit and function, and the failing dimension was never measured. That alone tells you the prototype did not prove it.
- Compare process routes. Ask the supplier for the routing sheet and compare the number of setups, the datum used in each, the fixturing and the stock form with how the prototype was made.
- Check the measurement before the part. Measure a few rejected parts by both methods, and if the disagreement persists, run a gauge study on that characteristic as described in the AIAG Measurement Systems Analysis manual.
- Look at the spread, not only the pass rate. Measure consecutive parts on the failing characteristic, in machining order, with the fixture, machine and material lot noted for each. A distribution centered in tolerance but too wide points to capability; one shifted toward a limit points to a setup, offset or tool wear; a steady drift along the run points to tool wear or heat; two separate groups usually trace back to two fixtures, two machines or two material lots, which the notes then confirm.
- Make one change and re-measure. Correct the cause you found, run a short batch on the production route and measure the same characteristic again before releasing the rest of the lot. If the fix touched the drawing, the process or the tooling, record it as a partial first article rather than an informal adjustment.
| Symptom in the production lot | Most likely cause | How to confirm | Typical fix |
|---|---|---|---|
| Position or perpendicularity between features from different setups fails | Relocation error between operations; datum not stated or not used for fixturing | Compare the routing with the prototype; check which datum each fixture locates on | Fixture on the drawing datums, combine operations, or add a functional datum to the drawing |
| Parts in tolerance but will not assemble | Prototype was fitted to its mate; tolerance stack never checked | Stack-up of the fit at worst case with both parts’ tolerances | Re-tolerance the fit, or add a functional gauge for the assembly feature |
| Threads or bores fail only after finishing | Finish buildup not allowed for; prototype was bare | Measure before and after finishing on the same part | State dimensions as after finish, add masking, or pre-size for buildup |
| Most parts good, a few out, wide spread | Tolerance near the limit of the process capability | Measure consecutive parts and compute the spread against the limits | Open the tolerance where function allows, or change the process for that feature |
| Whole lot shifted toward one limit | Offset, tool wear or setup error | Plot results in machining order | Correct the setup; add an in-process check on that feature |
| Thin walls or flat faces distorted | Different stock form or blank, residual stress, clamping | Compare material certificates and stock form with the prototype | Specify stock form or stress relief; rough, release, then finish |
| Supplier says pass, receiving says fail | Different measurement method or datum setup | Measure the same parts by both methods; gauge study | Agree one method and datum setup per characteristic in writing |
What has to change between prototype and production, and how do you decide?
Freeze the drawing, the material and finish, the process route and fixturing, and the inspection plan, then prove them together with a first article made by the production process. Then treat any later change to those items as a trigger for a new full or partial first article.
First article inspection under SAE AS9102 exists to verify that the production process, not a special prototype route, makes a part that meets every requirement on the drawing and purchase order, and it calls for a full or partial (delta) first article when the design, the manufacturing process or location, the tooling or the materials change. A first article made on soft jaws, another machine or prototype fixtures is evidence for that route only. The same idea runs through the AIAG PPAP manual (4th edition), which takes submission parts from a significant production run: one to eight hours of production and at least 300 consecutive parts, unless the customer specifies otherwise. Most non-automotive buyers do not need a full PPAP, but they do need the principle: approve the process that will make the parts, at the rate it will make them. The first article guide covers the report itself, and the PPAP guide covers when a fuller package is worth asking for.
Decide each change item by asking one question: if this changed tomorrow without telling you, could a part still pass inspection and fail in the product? If yes, it belongs on the drawing or the purchase order and in the change-notification list. Process route and fixturing almost always pass that test for features that relate two setups. Stock form passes it for thin or long parts. Finish passes it for every threaded, fitted or sealing feature.
| Item | Typical prototype practice | What production needs | Evidence to ask for |
|---|---|---|---|
| Drawing | Model plus a sketch; tolerances implied or verbal | Released revision, every dimension toleranced or covered by a general tolerance, datums from function | Marked-up drawing review from the supplier |
| Critical characteristics | Known to the designer only | Flagged on the drawing with how they are measured | Inspection plan listing them |
| Material and stock form | Whatever was on the shelf | Specification, temper or condition, and stock form if it matters | Material certificate per lot |
| Process route | One-off setups, probing and recutting | Fixed routing, operations and machines | Routing sheet |
| Fixturing | Vise, soft jaws, clamps | Dedicated fixtures locating on drawing datums | Fixture or tooling report |
| Finish | Bare or cosmetic only | Finish specification, masking, before or after finish dimensions | Finish certificate; dimensions after finish |
| Inspection | Calipers, fit check | Agreed method, gauges and sampling per characteristic | First article report; lot inspection record |
| Change control | None | Supplier notifies before changing site, process, tooling or material source | Clause on the purchase order |
What should the drawing, RFQ and purchase order say so production parts match the approved prototype?
Clean up the drawing so it defines the part without the prototype maker’s judgment, then state quantities, first article and change notification on the RFQ and purchase order. Anything the prototype shop decided informally has to become a written requirement or it will be decided again, differently.
Drawing cleanup.
- Tolerance every dimension or cite a general tolerance class, for example the medium class of ISO 2768-1, and loosen anything the prototype proved does not matter.
- Set datums from how the part is located in the assembly, and use them for every position and orientation tolerance, so the supplier can fixture on them.
- Flag the critical characteristics and say how each is checked; the critical characteristics guide shows the symbols.
- Replace “match mating part” or “fit to assembly” notes with a toleranced fit or a functional gauge.
- Name the material by specification and condition, and the finish by specification, with masking and a note that says whether dimensions apply before or after finishing.
- Release a new revision, and make sure the model and the drawing agree; if they disagree, say which one governs.
RFQ and purchase order.
- Quantities for each stage: prototype or pilot, first production release and annual volume, so the supplier quotes the right process route and fixtures.
- One-time items listed separately: fixtures, programs, gauges and the first article.
- A first article on the first production lot, made with production tooling and routing, measured against every ballooned characteristic.
- Capability on critical characteristics where it matters, stated as an index and sample size; the Cpk vs Ppk guide explains how to write that requirement.
- A change-notification clause: the supplier tells you before changing production site, process route, tooling, material source or finish source, and a new full or partial first article follows.
- The agreed measurement method for any characteristic where prototype and production checks differed.
Who does what on a Saola order when a prototype moves to production?
CNC turning and milling run at our lead CNC plant, T.K Viet Nam Precision Company Limited. Casting, stamping and all surface finishing go to partner plants we manage, and every quote names the production site and the country of origin of each part.
When a prototype drawing comes in for a production quote, the first step is a drawing review: you receive a marked-up copy of your own drawing and a short manufacturability note listing tolerances, datums or notes that a production route will read differently from a prototype shop. Machined features are made by CNC turning and CNC milling at the lead CNC plant, on fixtures built to your drawing datums. If the part needs anodizing, plating or another coating, that work is placed with a partner plant through our surface finishing program, and the quote says where it is done.
The first piece from production tooling is measured against every toleranced dimension, and you review and sign off the first article report before production continues. Production then runs to the routing verified at first article, and dimensional inspection records for each lot can be sent with the shipment. Material certification is available on request, and ISO 9001 certification is in progress, not held. If your program needs PPAP elements or capability data on specific characteristics, put them on the RFQ so they are quoted rather than discovered. Send the drawing, the model and your quantities through the quote form.
Frequently asked
Can we skip the first article if the prototypes already passed?
Usually not. The prototypes proved the design; the first article proves the production process, fixtures and measurement. If the prototypes were made by the same route, tooling and inspection as production, say so in writing and the first article may be reduced, but that is rare.
Should production run on the same machine as the prototype?
Not necessarily. What matters is that the production route is fixed, locates on the drawing datums and is proven by its own first article. A route that copies a prototype setup can be slow and expensive at volume.
Why is the production price not much lower than the prototype price?
Setup and programming spread over more parts, but production adds fixtures, gauges, a first article, finishing and documents, and prices every tolerance as written. If a drawing still carries prototype-era tight tolerances everywhere, opening the ones that do not matter is usually the largest lever.
Is a pilot lot worth ordering between prototype and production?
Often, when the part has features that relate several setups, a finish that affects fits, or a tolerance close to the process limit. A pilot lot made on the production route shows the spread and the assembly behavior before the full release is committed.
Who is responsible when a build-to-print part fails but the prototype was accepted?
On a build-to-print order the supplier is responsible for meeting the drawing as written, and the buyer for what the drawing says. If the drawing left the failing feature undefined, fix the drawing first, then agree how the parts already made are dispositioned.
Sources
- SAE AS9102C, Aerospace Series – First Article Inspection Requirements, SAE International, 2023
- Production Part Approval Process (PPAP), 4th edition, Automotive Industry Action Group, 2006
- SAE AS9145, Requirements for Advanced Product Quality Planning and Production Part Approval Process, SAE International, 2016
- ASME Y14.5-2018, Dimensioning and Tolerancing, American Society of Mechanical Engineers, 2018
- ISO 2768-1:1989, General tolerances – Part 1: Tolerances for linear and angular dimensions without individual tolerance indications, ISO, 1989
- NIST/SEMATECH e-Handbook of Statistical Methods, 6.1.6 What is Process Capability?, National Institute of Standards and Technology, 2012
- Measurement Systems Analysis (MSA), 4th edition, Automotive Industry Action Group, 2010